基于故障风险学习的无人机安全控制方法

In light of the fact that the safe maneuvering capability of unmanned aerial vehicles (UAVs) can significantly degrade under the influence of fault risks, an adaptive safety control approach grounded in fault risk quantification and learning is presented in this paper. First, building upon the fixed-time fault observation design, the conditional value-at-risk (CVaR) is adopted to quantitatively evaluate fault risks. Real-world experimental data are collected to examine the distribution of UAV positional uncertainty across varying fault risk levels, thereby facilitating explicit perception of fault risk states and effective capture of tail-risk events. Second, a lightweight backpropagation neural network combined with a sliding time window is leveraged to model the impact of fault risks on positional uncertainty, enabling swift risk perception and response. Furthermore, an adaptive risk-tendency control compensation strategy is incorporated to achieve smooth adjustment of control policies and realize adaptive safety control. Finally, experimental results verify that, compared with the integral sliding mode control method, the proposed adaptive control strategy, while imposing minimal computational overhead, enhances controller responsiveness and trajectory tracking accuracy under various fault conditions.